Block time-domain channel estimation in an OFDM system

Abstract

OFDM receiver that determines a time-domain channel impulse response, the receiver comprising: a filter (620) receiving OFDM symbols (610) containing data information and actual pilot information (520, 540, 560), the filter (620) being configured to comb filter and puncture the received OFDM symbols (610) to provide punctured OFDM symbols comprising the actual pilot information (520, 540, 560) and in which data information has been punctured; a memory (630) for receiving and storing the punctured OFDM symbols; a generator (650) coupled to the memory for generating additional pilot information (510) for the OFDM symbols based on the actual pilot information (520, 540, 560) of one or more of the punctured OFDM symbols; a time-domain channel estimator (670) that processes a first punctured OFDM symbol comprising the additional pilot information (510) to generate a channel impulse response for the first punctured OFDM symbol; and a frequency equalizer (690) that equalizes the first punctured OFDM symbol based on the channel impulse response for the first punctured OFDM symbol.

Description

BACKGROUND 1. Technical field

[0001] The present invention relates to a system and method for processing communication signals to achieve channel estimation more efficiently, particularly in providing channel estimation in an orthogonal frequency division multiplexing (OFDM) receiver. 2. State of the art

[0002] To increase data rates and mitigate multipaths, advanced networks, including so-called 4G wireless networks such as WiMAX and LTE (Long-Term Evolution), have adopted variants of the orthogonal frequency division multiplexing, or OFDM, signaling format for the PHY layer. The PHY layer is the physical electromagnetic means by which bits of information are sent and received over the air or wire. OFDM offers much-desired bandwidth efficiency with built-in mitigation for the multipaths of wireless channels in urban environments. The sensitivities of OFDM transmission are well understood.

[0003] The PHY layer bit pumping scheme has proven successful in digital subscriber line (DSL, wired) OFDM applications. Mobile wireless OFDM applications, however, still face challenges in achieving the designed capacity of OFDM.

[0004] At the core of the practical and theoretical advantages of OFDM lies the use of a fast Fourier transform (FFT). The FFT implemented in OFDM can be considered analogous to a bank of tuners for N c-simultaneous radio stations because each of the tones generated by the FFT can be independently assigned to users. The OFDM PHY delivers or receives a simultaneous burst of bits over a short period of time on each carrier frequency (tone), with a full or partial allocation of carriers to a given user. Partial allocation of carriers between different users and aggregating many users in one period is a multiple access scheme for OFDM. In the case of channels with 10 MHz bandwidth, a user can access up to N c = 840 (WiMAX) or 600 (LTE) simultaneous tones over a very short duration, such as 0.1 milliseconds. This N c - Tones per period form an OFDM symbol. The allocation of many users in one symbol is called OFDMA.

[0005] Wireless standards typically consist of three key time segments defined by the available bandwidth and time sensitivity of the information. Symbols are concatenated to define a frame, which is the longest relevant time unit and might be, for example, one millisecond. If the standards assign ten symbols to a frame, the symbol duration is 0.1 milliseconds. Finally, the FFT size and the cyclic prefix (CP) duration define the time spacing between samples, so a 1024-point FFT and a 128-point CP define a sampling interval of 11 microseconds. Although FFT computations can be comparatively efficient, the FFT size for an exemplary OFDM system is large enough (e.g., 1024 samples in the case of 10 MHz bandwidth) that the computational requirements remain relatively high, and power consumption remains an important constraint in the design of receivers for user handsets.

[0006] OFDM systems are more sensitive and have less robust signal acquisition than 3G systems based on CDMA (Code Division Multiple Access). The sensitivity of OFDM systems stems from their use of the fast Fourier transform (FFT) to transform incoming signals from the time to the frequency domain. The FFT in OFDM systems can deviate from ideal assumptions under very common real-world conditions and receiver implementations. If the assumptions underlying the FFT algorithm fail, crosstalk develops between all N c channels (on N c carriers) being transmitted. Crosstalk between carriers degrades performance, which in turn causes increases in bit error rates (BER).

[0007] An OFDM wireless handset may receive multiple paths (copies with different delays) of the same signal from a transmission tower ("base station") due to reflections from structures or large bodies of water. This non-line-of-sight, or multipath, causes the signal to be distorted from the linear frequency-domain "shape" output by the transmitter. A receiver must compute a filter to restore the signal to its original linear spectral shape; this filter is called equalizing the signal. OFDM receivers perform a critical equalization calculation for each transmitted OFDM symbol.

[0008] Unlike most other modulation strategies commonly used in communications systems, OFDM can incorporate two equalizers to improve signal quality: a timing equalizer (TEQ) and a frequency equalizer (FEQ). Some OFDM applications, such as DSL, include a timing equalizer, while others, such as systems implementing current wireless standards, do not require a timing equalizer. All practical OFDM receivers include a frequency equalizer. Whether a receiver includes a timing equalizer or only a frequency equalizer, the receiver must perform channel estimation to determine, at least initially, values of the equalizer coefficients before the equalizer can be used to improve signal quality. The determination of coefficients for frequency equalizers is typically done in the frequency domain.

[0009] An OFDM communication system typically includes an OFDM transmitter that generates radio signals modulated with information, such as data generated by a computer network or voice data. The radio signals propagate to a receiver over a channel that distorts the radio signal in various ways, such as transmitting it over multiple paths of varying lengths, thereby introducing multiple copies of the radio signal with different offsets and amplitudes in a mechanism known as multipath. Receiver circuitry downconverts the received signal to baseband and converts this signal from analog to digital to produce the information signal, which undergoes OFDM processing. The radio signal is time-aligned. After alignment, the signal is processed to remove the cyclic prefix (CP) from the signal. The cyclic prefix is present because OFDM transmitters use a CP of length N CP, which consists of the last N CP samples to an information signal of length N, so that the digital signal that the transmitter converts to analog and sends has length N + N CP An initial step of the receiver's reverse conversion process is then the removal and discarding of the added N CP Cycle prefix samples. After this step, the serial signal is organized by a parallel conversion element and converted to a parallel signal for further processing. The cyclic prefix can be removed either before or after the serial-to-parallel conversion.

[0010] After CP removal, the parallel data is fed to a fast Fourier transform (FFT) processor, which converts the time domain samples s(n) into a set of frequency domain samples R i(k). The received OFDM symbol is assumed to be corrupted by the channel, which, for OFDM, is assumed to introduce amplitude and phase distortions of the samples from each of the carrier frequencies used in the OFDM system. A frequency equalizer (FEQ) applies an amplitude and phase correction to the different samples transmitted at the different frequencies, specific to each of the frequencies used in the OFDM system. The FEQ requires an estimate of the channel's amplitude and phase deviations from the ideal case at each frequency to determine which corrections to apply.

[0011] A typical OFDM channel estimator receives and estimates a channel in the frequency domain based on a set of pilot tone locations and received pilot signals. This is referred to as frequency-domain channel estimation, or FDCE. The pilot tones (or simply pilots) are typically one- or two-bit symbols prescribed by the relevant standards, so the receiver knows the expected pilot locations and values a priori. All FDCE implementations respond to the OFDM symbol output by the FFT to extract the received pilot signals. The channel estimate at each pilot can be determined as the amplitude and phase rotation from the ideally expected post-demodulation value of "+1" for each pilot. Any deviation from this value of "+1" represents the distortion from the channel at that frequency's bandwidth.The value of the channel at the data carrier frequencies can be estimated by interpolating the values obtained at the pilot carrier frequencies. Various improvements to simple channel estimation methods are known and are conventionally implemented in the frequency domain. The frequency equalizer receives the signals from the fast Fourier transform processor and channel estimates from the estimator and equalizes the signal. The output of the equalizer is typically fed to a parallel-to-serial element, which converts the parallel outputs of the equalizer into a serial output user signal.

[0012] An OFDM symbol is constructed by setting active carrier values to non-zero values from a prescribed set of values according to the number of bits to be "loaded" into that OFDM symbol. These values are then subjected to an inverse fast Fourier transform (IFFT) to obtain the time-domain samples. The cyclic prefix is added to the beginning of the symbol by taking a defined number of samples from the end of a symbol's sequence of time-domain samples. The IFFT could, for example, produce 1024 samples. Certain standards select the CP to have a length of 128. That is, the transmitter selects the last 128 samples from the sequence of 1024 samples and prepends these samples to become the first 128 samples in the transmitted OFDM symbol, which has a total of 1152 samples.Due to this construction, selecting any 1024 samples from the 1152 samples of the OFDM symbol produces a circular shift on the original 1024 OFDM time domain samples.

[0013] In the case of the WiMAX standard, the OFDM symbol can be transmitted on 60 subchannels with 14 active carriers per subchannel for a total of 840 active carriers, with 4 pilots per subchannel. The locations of the pilots within any given symbol, and therefore subchannel, are prescribed by the standard. OFDM methods for high-throughput networks attempt to minimize overhead, and this includes the number of training carriers in a symbol. Reducing the number or density of pilots can limit the ability of receivers to efficiently recover information from a signal.

[0014] A theoretical advantage of OFDM is that post-FFT equalization can be performed for each received tone individually using a relatively simple algorithm. Another advantage enabling OFDM receivers is that equalization coefficients only need to be estimated for each subcarrier relevant to the user, a quantity smaller than the FFT size. The values for each equalization coefficient corresponding to each tone will depend on the estimate of the channel coefficient—referred to as channel estimation. Like many operations in OFDM receivers, typical OFDM receivers perform post-FFT channel estimation because channel estimation is easily and efficiently performed at this point based on a user's tone allocation. Because channel estimation is performed post-FFT, the tones are affected by FFT and post-FFT distortions, referred to as intercarrier interference (ICI).ICI generally manifests itself through three conditions: 1) frequency tuning errors; 2) Doppler effects from mobility; and 3) interference from other cell sites. OFDM systems account for intersymbol interference by providing a time gap between symbols, so intersymbol interference is generally less problematic for OFDM compared to other wireless methods.

[0015] Any given channel has a well-known limit to its capacity. In current OFDM implementations, there is additional capacity loss below the expected rates. Channel estimation errors are a major culprit. Because ICI affects the channel estimation algorithms after the FFT in typical implementations, poor channel estimation leads to inaccurate equalizer coefficients. Increased bit error rate (BER) due to myriad conditions, such as demanding channels and poor channel estimation, can be accommodated by reducing the transmitted bit rate offered to a user. Effectively, reducing the transmitted bit rate allows robustness against interference.However, this is a non-linear correction, since the OFDM method allows transmission of two, four or six bits per tone and consequently, under certain circumstances, distortion reduction requires sending less than 2 bits per tone, which means that the system provides no data to the user at all.

[0016] US Patent US 8,385,438 B1 discloses a system and method for adaptively synchronizing a transform window in a multi-subcarrier communication system based on an analysis of a time-domain channel impulse response. Patent application US 2011 / 0249773 A1 discloses a method for estimating a channel in a wireless communication system. The method includes determining channel estimates of pilot tones, selecting data tones to which a first estimation scheme is applied according to the frequency selectivity and time-axis variance of the channel, determining channel estimates of the selected data tones according to the first estimation scheme by using the channel estimates of the pilot tones, and determining channel estimates of the remaining data tones according to a second estimation scheme using the channel estimates of the pilot tones and the channel estimates determined by the first estimation scheme. SUMMARY OF THE PREFERRED EMBODIMENTS

[0017] The invention is the subject of independent patent claim 1. Advantageous embodiments of the invention are the subject of the dependent patent claims.

[0018] One aspect of the present invention provides an OFDM receiver that determines a time-domain channel impulse response. The receiver includes a filter that receives OFDM symbols and performs comb filtering and puncturing of OFDM symbols to provide punctured OFDM symbols including pilot information. Punctured OFDM symbol storage receives and stores a predetermined number of punctured OFDM symbols. Coupled to the punctured OFDM symbol storage is a generator for generating virtual (additional) pilot information that is introduced into OFDM symbols. A time-domain channel estimator processes a first OFDM symbol including virtual pilot information to generate a channel impulse response for the first OFDM symbol. A frequency equalizer equalizes the OFDM symbol in response to the channel impulse response for the first OFDM symbol.

[0019] According to another aspect of the present embodiment, the receiver may perform interpolation to determine the pilot values. The receiver may use a Wiener filter to perform the interpolation and may use Doppler information and / or the signal-to-noise ratio in determining the pilot values.

[0020] According to another further aspect of the present embodiment, the generator generates virtual pilot values for data carrier locations using values of channel samples associated with pilot locations. The receiver further includes a CIR selector responsive to an initial channel impulse response to select a channel impulse response interval, wherein the time-domain channel estimator responsive to the channel impulse response interval to generate another channel impulse response. In another aspect, the CIR selector identifies channel paths for inclusion in the channel impulse response interval, and the CIR selector terminates a process of identifying channel paths based on a normalized peak value associated with a last identified channel path.

[0021] Another aspect of the invention provides a method for communicating in an OFDM communication system, comprising receiving and filtering OFDM symbols to provide punctured OFDM symbols having sampled pilot information sampled from the channel over which the OFDM symbols were transmitted. Virtual pilot values corresponding to a plurality of data carrier locations are generated. The virtual pilot values are determined from the sampled pilot information. Using the virtual pilot values and the sampled pilot information, a channel impulse response corresponding to a first OFDM symbol is estimated. The first OFDM symbol is equalized in response to the channel impulse response for the first OFDM symbol.

[0022] Another aspect of the present embodiment may be determining an initial channel impulse response, wherein the resulting initial channel impulse response is a time-domain channel estimate. In another aspect of this embodiment, a process of selecting a channel impulse response comprises selecting a channel impulse response interval. A process of determining the channel impulse response interval may identify channel paths to be included in the channel impulse response interval. Further, a process of identifying channel paths is stopped based on a normalized peak value associated with a most recently identified channel path. BRIEF DESCRIPTION OF THE DRAWINGS Fig. Figure 1 schematically shows an OFDM communication system with an OFDM transmitter and a time-domain channel estimation OFDM receiver. Fig. 2 shows an initial channel estimate, a final channel estimate, and an idealized channel representation. Fig. Figure 3 schematically shows the distribution of pilot carriers and data carriers in a fourteen-symbol subframe according to the LTE (Long-Term Evolution) standard. Fig. Figure 4 shows schematically the distribution of pilot carriers and data carriers of Fig. 3 after a puncturing operation, which, among other things, can zero or set the disks in a subframe to zero. Fig. Figure 5 shows a subframe being processed to provide virtual pilots to enable time-domain channel estimation. Fig. 5 can be generated, for example, by inserting virtual pilots into the subframes of Fig. 4, which has undergone a puncturing process. Fig. Figure 6 shows circuits for generating virtual pilot information and using this information in time-domain channel estimation. Fig. Figure 7 shows aspects of a time-domain channel estimation process with respect to an initial channel estimate. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] A typical OFDM receiver includes a simple frequency-domain channel estimator (FDCE) functionally colocated with the frequency equalizer. The FDCE uses the frequency representation of the OFDM symbol and possibly other information to estimate the channel symbol-by-symbol or block-symbol-by-block. The equalizer coefficients are individual complex-valued weights for each active carrier. In multiple-access methods, the carriers dedicated to a single user are typically a subset considerably smaller than all active carriers. For example, in an LTE configuration, 600 downlink carriers are available to a user, but as few as 36 may be allocated to that user. The channel estimation can therefore be effectively computed in groups of 12 adjacent frequencies. In a multiple-access OFDM implementation, estimating the channel only in the receiver for a user can provide a useful simplification.

[0024] Time-domain channel estimation (TDCE) offers significant advantages over frequency-domain channel estimation (FDCE) in complex network designs or for high user densities. The increased robustness provided by time-domain channel estimation translates into higher network-wide throughput. The original Wi-Fi standard (802.11a), which uses OFDM, cannot realize performance gains from implementing time-domain channel estimation, but uses of TDCE in receivers implementing the comparatively complex LTE (Long-Term Evolution) standard can achieve greatly improved throughput and an enhanced user experience.

[0025] One problem with time-domain channel estimation is that there may be too few pilot carriers in any given symbol to perform effective channel estimation. In a block symbol transmission scheme, there must be a minimum pilot density to enable time-domain channel estimation. Standards such as LTE are designed for frequency-domain channel estimation and do not necessarily provide enough pilot signals in a symbol for time-domain channel estimation. Not all symbols in a given block have pilot carriers, and the pilot density in pilot-carrying symbols may be insufficient for robust time-domain channel estimation.

[0026] Preferred aspects of the present invention can provide robust time-domain channel estimation, for example, by increasing the pilot density for block symbol OFDM transmission systems such as LTE. LTE typically provides fourteen symbols in any given group, and pilots are typically included in four of these symbols. The aggregation of fourteen symbols is referred to as a subframe. The subframe duration is 1 ms, and each carrier spans 15 kHz. An LTE system using a 10 MHz bandwidth channel may use a 1024-point FFT with 600 active carriers. It follows that the 1 ms transmission distributes the necessary information for transmission from the base station to the mobile user into as many as 600 x 14 = 6240 carrier intervals. These intervals are referred to as "resource elements" in LTE.In certain LTE configurations, there are too few pilots in the symbols that comprise the subframe, which can lead to convergence and estimation accuracy problems, among other issues. Preferred implementations of the present invention provide the effective equivalent of a number of additional pilots (virtual pilots), which can significantly improve the feasibility and performance of time-domain channel estimation. Additionally, channel impulse response processing can provide a better framework for channel estimation, thereby also improving the feasibility and performance of time-domain channel estimation. The following discussion provides an overview of a suitable receiver suitable for implementing and exploiting time-domain channel estimation.

[0027] Fig. Figure 1 highlights the functionality of a time-domain channel estimation receiver for symbol-wise processing in an OFDM communication system. The OFDM communication system of Fig. 1 includes an OFDM transmitter 10 that generates radio signals modulated with information such as data generated by a computer network or voice data. The radio signal propagates over channel 12 to a preferred implementation of a TDCE-OFDM receiver. Channel 12 distorts the radio signal in various ways, including transmitting it over multiple paths of varying lengths, thereby introducing multiple copies of the radio signal with different offsets and amplitudes in the mechanism known as multipathing. The radio signal is downconverted and input to an alignment element 14, which time-aligns the signal so that it can be processed according to transmission standards. After alignment, the data is passed to a processing element 16, which removes the cycle prefix (CP) from the signal.After this step, the serial signal is organized by a serial-to-parallel conversion element and converted into a parallel arrangement for further processing. The cyclic prefix can be removed either before or after the serial-to-parallel conversion.

[0028] After CP removal 16, the parallel data is fed to a fast Fourier transform (FFT) processor 18, which converts the time-domain samples s(n) into a set of frequency-domain samples Ri(k) for processing. The received OFDM symbol is assumed to be corrupted by the channel, which, for OFDM, is assumed to introduce amplitude and phase distortions of the values at each of the subcarrier frequencies used in the OFDM system. A frequency equalizer 150 may apply specific amplitude and phase correction for each of the subcarrier frequencies used in the OFDM system to the various samples transmitted at the different frequencies. The correction applied by the FEQ 150 preferably uses a channel estimate of the amplitude and phase deviations of the channel from the ideal case, with the channel estimate preferably being provided in the time domain. Certain preferred implementations of the TDCE receiver of Fig. 1 determine a channel estimate for each received OFDM symbol. Other preferred implementations use statistical measures to provide robust functionality against known impairments. The equalized symbol output by frequency equalizer 150 is fed to decoder 20, which processes the symbol to extract the transmitted data.

[0029] The pilot location element 22 stores and outputs a set of pilot signal locations and modulation values according to the considered standard. The pilot location element 22 can output pilot signal locations corresponding to the symbols and subcarriers that the corresponding communications standard prescribes as pilot signals. Optionally, the pilot location element 22 also outputs virtual pilot locations in addition to the pilot signal locations and values prescribed by, and preferably generated from, the standards. The additional virtual pilot signals provide increased pilot signal stimulus, allowing responsive elements to generate more accurate outputs, which can provide greater stability.The reference signal element 24 is preferably responsive to pilot location information output by the pilot location element 22, and is preferably responsive to the actual and virtual pilot locations and modulations to generate a reference signal with augmented pilot signal location stimulation. In some implementations, the pilot location element 22 outputs, for each actual and virtual pilot location, phase and amplitude information associated with actual and virtual pilot signal locations in the frequency domain. Other circuitry, such as the reference signal element 24, could provide one or more of these data sets, depending on how the circuitry is implemented and the complexity of the implementation, or one or more of these data sets might not be needed in certain implementations. The reference signal generated by the element 24 can be a time-domain signal or a frequency-domain signal, as desired.The reference signal output by element 24 can be selected with respect to correlations between the reference signal and either a time domain or frequency domain receive signal.

[0030] An approach used in the OFDM receiver of Fig. 1 is to obtain an initial estimate of the time-domain channel impulse response (CIR), which is advantageously determined by a statistical measure. This statistical measure is typically the correlation between a reference signal provided by module 24 and the received symbol. This calculation may be performed in a statistical measure module 140, which performs a preferred correlation between the reference signal and the received signal to generate an initial channel impulse response. Two additional preferred processes remain: selecting the time period containing the significant channel paths, preferably derived from the initial estimate, and estimating the channel to pass the channel estimate to the equalizer. The initial channel impulse response is output to the CIR selection module 120 to select the time period that captures the significant channel paths.

[0031] The CIR selection module 120 is configured to process a vector of values from an initial channel estimate (initial CIR) and output a shorter vector containing an estimate of the time window containing the significant paths in the channel. In general, the CIR selection module is designed to select the best window that encompasses the information about the significant paths without capturing unwanted noise or requiring the processing of an undesirable number of samples. Fig. 2, curve 240, shows an initial estimate for a three-way channel. The initial CIR waveform 240 has the Fig. 2 initial CIR duration specified at 270, which consists of M samples arranged in a vector. The true channel impulse response (CIR) is in Fig. 2 as three paths 222, 224, and 226. A receiver can preferably be programmed to set the value for the number M of samples comprising the initial time-domain channel estimate 240. Preferably, the number M of samples is chosen to exceed the target CIR duration 260 of L samples. That is, the CIR selection module 120 preferably determines which L samples comprising the CIR duration 260 are to be selected from the M samples in the initial CIR duration 270, where L < M. Preferably, the CIR selection module 120 selects the L samples that best represent the CIR at which the significant paths are present in the channel. Typically, M does not exceed the number of samples in an OFDM symbol.

[0032] Fig. Figure 2 shows three curves of interest that determine the robustness and performance of a preferred OFDM receiver implementing time-domain channel estimation. The first curve is the initial channel estimate 240, the second curve is the achievable channel estimate 250, and the third curve is the set of three paths 222, 224, and 226 that comprise the actual channel. The estimate of curve 240, when compared to simple estimators of the channel frequency response (CFR), is insufficient to achieve a target bit error rate. The correlation performed by the statistical measure module 140 to determine an initial estimate does not exhibit near-orthogonal properties in its autocorrelation response at delays other than zero.This contrasts with other transmission methods that use orthogonal codes, such as CDMA (Code Division Multiple Access) in the WCDMA standard, which uses spread spectrum theory and a set of orthogonal codes. Indeed, OFDM is characterized by relatively poor autocorrelation properties, as demonstrated by the "broad" peaks shown in the initial channel estimate curve 240. For theoretically ideal CDMA spreading codes and lengths, this same correlation for generating the initial channel estimate can be closer to that of curve 250.

[0033] The CIR selection module 120 preferably uses the initial CIR estimation response 240 to select the L samples for the preferred CIR duration from the M samples that comprise the initial CIR 240 and that span the initial CIR duration 270. The estimation module 130 preferably uses this initial CIR to determine a best CIR estimate. Suitable channel estimators are described, for example, in US patent application US 2013 / 0121392 A1. The estimation module 130 is capable of "removing" the non-orthogonal correlation properties of an OFDM symbol. The output of the estimator 130 is shown in Fig. 2 as the estimated CIR curve 250. This CIR estimate 250 performs better than typical OFDM receivers with FDCE implementations under conditions that do not violate the FDCE's assumptions about the OFDM symbol and channel conditions. One such basic assumption is that the actual channel duration, the delay period between the first 222 and farthest significant path 226, does not exceed the length of the cyclic prefix. Violation of the cyclic prefix duration constraint, while retroactive for FDCE, is of significantly less impact for the time-domain channel estimation (TDCE) methods preferred here. Provided that the statistical measure module 140 accounts for this potential condition, the CIR selection 120 and the estimator 130 are capable of providing a highly accurate estimate of the true channel.This violation of the cyclic prefix duration constraint is expected to occur in certain network configurations for LTE deployments in the near future.

[0034] In certain preferred embodiments of an OFDM receiver, the CIR selector 120 preferably selects a portion of the initial CIR for further processing to develop a channel estimate, or may otherwise achieve a channel estimate with a length shorter than the symbol length or the length of the initial CIR. Such preferred embodiments may, for example, utilize metrics characterizing the channel to advantageously determine a shortening of the initial CIR that is useful in terms of complexity, robustness, and accuracy for time-domain channel estimation. A suitable metric for evaluating the CIR duration could, for example, be generated by the iteration controller 26 or could be generated by another element of the receiver from Fig. 1 in response to the channel.

[0035] Preferred embodiments of the receiver of Fig. 1 include an iteration controller 26 for controlling the number of iterations of the channel estimator 130. The iteration controller 26 preferably receives and considers information from one or more of the CIR selector 120, which outputs an initial channel estimate, the statistical measure element 140, and the channel estimator 130. The iteration controller 26 and the channel estimator 130 preferably cooperate to achieve desired channel estimation performance under varying pilot configurations during training and / or pilot / data mixed symbols. The iteration controller 26 is advantageous for channel estimation in the channel estimator 130 over computational methods that utilize second-order moment measures, such as correlation or autocovariance matrices. Direct calculations for these formulations are comparatively less likely to be numerically stable, and thus iterative methods are preferably used.For this reason, metrics and other quality measures are preferably processed to determine a sufficient number of iterations or, for example, a maximum number of such iterations.

[0036] Under many circumstances, the channel estimator 130 does not provide a CIR that is properly aligned for equalization. Preferably, the phase alignment module 28 then responds to metrics from the iteration control module 26 to properly adjust the CIR to match the frequency-domain phase of the corresponding OFDM symbol being processed by the TDCE receiver. After phase alignment, the channel estimate is extended or stuffed to have a proper length for further processing. For example, the stuffing element 28 may insert trailing zeros so that the channel estimate has the proper length. Next, the fast Fourier transform element 30 transforms the time-domain channel estimate into the frequency domain for use by the frequency equalizer 150. Additional information about the structure, characteristics, and operation of the Fig. The circuits shown in Figure 1 can be found in the previously mentioned US patent application US 2013 / 0121392 A1.

[0037] Fig. Figure 1 shows a TDCE-OFDM receiver that assumes two essential operating conditions: (1) each processed symbol comprises a mixture of data and pilot carriers; and (2) in any given symbol, there are enough pilots to allow convergence of the estimation process 140. If not all symbols have pilots, then some type of averaging strategy 110 is preferably used to provide a channel estimate for data-only symbols (i.e., a symbol that contains no pilots and contains only data). These two essential operating conditions, however, may not exist in typical LTE configurations unless the virtual pilot strategy discussed below is introduced.

[0038] Fig. Figure 2 highlights certain problems with OFDM receivers using time-domain channel estimation. First, it is desirable for the receiver to accurately and precisely determine the CIR duration and which of the initial CIR samples are within the CIR duration. Second, it is desirable for the receiver to obtain an estimated CIR 250 that corresponds to the real channel in Fig. 2 as three paths (222, 224 and 226). The preferred receiver implements a method for the CIR selection 140 and estimation 120 modules to Fig. 2 when the OFDM symbols are arranged in blocks.

[0039] Aspects of the present invention provide advantageous implementations of time-domain channel estimation and are particularly applicable to block OFDM symbol systems. Implementations can provide high accuracy and robustness for realistic mobile environments, even at low pilot signal densities.

[0040] In LTE, the transmission of information bits to a user is segmented across a number of carriers and a number of consecutive symbols. While the LTE configuration may have a total of K carriers in use by a base station, the user may be allocated a number significantly smaller than K. The K total carriers in an LTE configuration are divided into groups of K RB divided into contiguous carriers. K RBis the number of carriers in a so-called resource block (RB), and a user can be allocated a number of non-contiguous RBs. For example, K = 600, KRB = 12, for a total of 50 RBs in a symbol, arranged along the frequency axis. Typically, LTE also features time-axis allocations, which are generally segmented into "subframes" of 1 ms duration, so that 14 OFDM symbols are present in each subframe. Ten subframes constitute a frame. Fig. Figure 3 shows a possible segmentation and arrangement of an LTE subframe in time and frequency with 14 symbols distributed over the time axis.

[0041] Fig. Figure 3 shows the pilot signal density in an LTE subframe, which apparently consists of three pilots in two symbols (No. 1 and No. 9) and two pilots in two other symbols (No. 4 and No. 13). Each set of pilots, identified by their representation 310 and 320, is located at different frequencies. This pattern may be repeated in subsequent subframes. In terms of pilot density, there are 10 pilots in a total of 168 active carriers, which can lead to convergence and accuracy problems, among other issues. Preferred implementations of the present invention provide the effective equivalent of a number of additional pilots (virtual pilots), which can significantly improve the feasibility and performance of time-domain channel estimation. Furthermore, the time-domain channel estimation receiver preferably uses an estimated CIR duration that closely matches the true CIR duration.The following discussion provides an overview of a suitable strategy for implementing and exploiting virtual pilots in time-domain channel estimation.

[0042] Comb Filtering and Puncturing To enable a preferred time-domain channel estimation receiver to identify the channel with the highest possible accuracy, the receiver preferentially increases the pilot density and preferably comb filters the received symbols to remove data carrier. The simplest of the comb filter implementations use an FFT. This is because the function of a comb filter is to decompose a signal into components, analogous to a filter bank, which can be readily implemented using an FFT or other transforms (e.g., DCT, wavelets, etc.). After the received symbol is transformed, the known data carrier intervals can be zeroed, for example, by zeroing the received amplitudes, to produce a punctured symbol that is more suitable for channel estimation.

[0043] Since the OFDM symbol is preferably converted from the time to the frequency domain, the receiver preferably applies the comb filter in the frequency domain to zero the data carriers. This zeroing of signal values at the data carrier locations can be referred to as "puncturing." Accordingly, the block of received symbols is divided into Fig. 3 preferably comb filtered to isolate the individual carriers, followed by zeroing or puncturing the data carriers to obtain the subframe and symbols, as in Fig. 4. In this representation, all data bit modulated carriers are set to a value of zero or punctured, as shown in the changes from the carrier and time index example at the location at 430. All pilot locations are effectively multiplied by 1 and all data locations are multiplied by zero, or the most efficient implementation to achieve a puncturing result, as shown in Fig. 4. It should be noted that carrier offset, Doppler effect due to mobility, and other received imperfections will cause crosstalk between adjacent channels. However, puncturing ignores these effects because channel estimation can reasonably ignore crosstalk between frequency carriers. Other strategies can be pursued that can enhance puncturing, for example, by filtering three contiguous carriers before puncturing. Such enhanced strategies may be desirable, particularly depending on the sensitivities of the OFDM communication system under consideration.

[0044] In certain situations, it is advantageous to multiply the data by zero, and the pilots by the conjugate of their known transmit values. For these situations, the pilot locations, after conjugate multiplying their transmit value, represent a value of the sampled channel at the pilot's frequency. The specific implementation of the statistical metrics module 140 or an equivalent module that produces an initial time-domain channel estimate determines the appropriate multiplier for the pilots in Fig. 4.

[0045] Virtual Pilots After comb filtering and puncturing of data carriers as in Fig. 4, each OFDM symbol is preferentially processed to increase the pilot density by introducing virtual pilots. The transmitted pilots have sampled the channel and have sufficient density for interpolation to other strategically chosen locations to effectively increase the pilot signal (tone) density in a given block of OFDM symbols. This increase in pilot density enables convergence of preferred time-domain channel estimation strategies, making these strategies capable of high estimation accuracy.

[0046] Various strategies are available to interpolate the channel estimate at the pilot locations to provide channel estimates at data carrier locations. A preferred implementation uses a two-dimensional Wiener filter implementation to estimate the virtual pilot values (phases and amplitudes) at the selected locations based on the measured pilot values and the standard-defined positions in the block. Alternatively, the interpolation can be more simply implemented with two-dimensional Wiener filters, which avoid estimating frequency-axis correlations and use only the Doppler and SNR estimates to perform a one-dimensional Wiener filter.If the receiver generates Doppler bandwidth and / or signal-to-noise ratio (SNR) estimates, the receiver can readily determine the autocovariance and cross-correlation vectors for the one-dimensional Wiener filter as a function of one or both of these variables, thus allowing a metric for Wiener filtering.

[0047] Fig. Figure 5 shows the use of known pilots in a two-dimensional grid of time and frequency of OFDM symbols to determine where to place virtual pilots at appropriate punctured data carrier locations. There are many variations for distributing virtual pilots at these data carrier locations, which are illustrated by the methods described above with reference to Fig. 4 described puncturing strategy method. Each zeroed location in Fig. 4, a virtual pilot can be assigned based on the locations and value of known pilots. In Fig. 5, all fourteen symbols at the five frequencies are assigned virtual pilots that correspond to the frequencies of the actual pilots in the received symbol.

[0048] Fig. Figure 5 shows in detail how a (complex) value can be assigned to the virtual pilot location 510 based on the interpolation from the nearest pilot locations for all sets in the OFDM block. That is, the value assigned to the virtual pilot location 510 is preferably achieved by interpolation from the nearest actual pilots for pilots in set no. 1 at locations 540 and 560 and for a pilot in set no. 2 at location 520. This procedure is repeated for each of the designated virtual pilot locations in Fig. 5. After determining the virtual pilot values, two results were obtained: The pilot density has increased from ~ 1 / 17 (in Fig. 4) to more than 1 / 3 (in Fig. 5) has increased; and there are now channel-sampled values for all symbols in the processed block of OFDM symbols.

[0049] Block OFDM symbol processing in TDCE A preferred OFDM receiver with a time-domain channel estimator is Fig. 6. The present inventors have tested the presented TDCE-OFDM receiver implementation using realistic simulations and observed that the receiver achieved close-to-theory performance under the adverse conditions assumed by known tests of CIR estimation strategies.

[0050] The present inventors have observed that sufficient pilot density in a block OFDM symbol and identification of a suitable CIR duration provide significant advantages for pragmatic implementations of TDCE-OFDM receivers. Having sufficient pilot density can determine the convergence or divergence of the CIR identification strategy in the time domain. Achieving a desirable CIR duration can determine receiver performance by ensuring that the identification process does not miss any significant path while reducing the complexity of time-domain channel estimation.

[0051] The recipient of Fig. 6 receives the OFDM symbols 610 with already established synchronization from the network. This helps determine the epochs that delimit the collection of subframe symbols. The network synchronization also allows a rough estimate for the beginning of each OFDM symbol. The comb filtering and puncturing module 620 implements the comb filtering and data carrier puncturing method described above to Fig. 4 once a subframe is processed. At the output of module 620, filtered and punctured OFDM symbols are stored in symbol queue module 630 until a predetermined number of symbols are in the queue. Pilot location module 640 provides temporary storage for pilot location data synchronized with the queued symbols in module 630.

[0052] Furthermore, in the frequency domain processing part of the Fig. 6, the virtual pilot interpolation module 650 responds to the queued OFDM symbols 630, which preferably have a dotted shape, such as in Fig. 4, and the information about the pilot locations 640 to generate virtual pilots. Preferably, the generation of the virtual pilots proceeds using nearest neighboring actual pilots, as shown in Fig. 5 and discussed above. Preferred interpolation strategies, such as two-dimensional Wiener filtering, can advantageously reduce the computational complexity of determining virtual pilot values using knowledge of certain symbol metrics, such as Doppler and SNR estimates. Preferably, the virtual pilot interpolation module 650 can then benefit from the estimation of the Doppler and SNR parameters. As discussed in more detail in the above-incorporated application Ser. No. 13 / 416,990, preferred embodiments of the receiver estimate Fig. 1 Doppler and SNR (SINR) metrics, and thus it is particularly useful to use these parameters when interpolating virtual pilots, for example, using two-dimensional Wiener filtering.

[0053] The receiver can determine Doppler and SNR metrics using strategies that depend on frequency and / or time-domain OFDM symbol representations. Fig. 6 provides a metrics calculation module 664 in the time domain processing portion of the receiver to determine metrics that are at least Doppler and SNR values. The metrics calculation 664 preferably acts on the CIR waveform in the CIR duration 260 output by the CIR selection module 662. The CIR selection module 662 in Fig. 6 is preferably the same as that in Fig. 1. These preferred metrics vary relatively slowly from symbol to symbol, so the effect of computational delay is insignificant. Thus, the virtual pilot interpolation module 650 preferably responds to values from the metrics calculation module 664, which is responsive to the waveform output from the CIR selection module 662 during the CIR duration 260.

[0054] The Wiener filter provides a particularly preferred interpolation strategy for generating amplitude and phase estimates of virtual zeroed data positions from the measured pilot symbol amplitudes and phases, especially when required for highest information throughput conditions. The Wiener-Hopf equation can determine a best unbiased estimate of an unknown parameter based on second-order statistics from statistical cross-correlation and autocorrelation measures.

[0055] The Wiener-Hopf equation has the form w=R−1p, where R is the autocovariance matrix and p is the cross-correlation vector. The weights that comprise the vector w are used to filter, or in this case, interpolate, the measured channel estimates to generate the desired estimates of the virtual pilots from the actual pilot locations and values. The values of R and p for such an interpolation can be estimated based on three parameters alone. Two of these three parameters, SNR and maximum Doppler frequency (f Dmax ) can be measured from the CIR estimate output by the channel estimation element 130. Preferably, the CIR selection element 662 is responsive to the channel estimation element 130 to provide the desired output of SNR and maximum Doppler frequency (f Dmax). The third parameter is determined by the location of the pilot-carrying symbols in the subframe, which in this LTE example consists of fourteen symbols. That is, for any given network-imposed receiver configuration, Δt is a static value.

[0056] Additional information on determining and storing the Doppler and SNR information, along with the general operation and implementation (albeit in a slightly different application) of a two-dimensional Wiener filter, can be found in the aforementioned US patent application US 2013 / 0121392 A1.

[0057] The waveform spanning the CIR duration 260 is the output of the CIR selection module 662 in Fig. 6. In the simplest implementation, module 662 identifies a range of preset time durations 260 that contains the most energy from the initial estimate waveform 240. The preset time duration is determined based on the operating conditions for the receiver, thereby tuning complexity as a function of the maximum channel delays in the physical geometries of the receiver environment.

[0058] A preferred implementation of the CIR selection module 662 includes strategies for selecting likely paths and establishing a time tolerance around those identified paths, from which the CIR duration 260 is selected. Fig. Figure 7 shows a preferred procedure to be implemented in the CIR selection module 662. The module 662 preferably identifies the significant paths and preferably reduces the duration of the initial channel estimation waveform 740 to the preset target duration 710. The module 662 preferably determines likely path peaks using known analytical tools. In the example of Fig. In Figure 7, the paths are denoted by an X, and a tolerance time window 733 is shown around the peaks associated with the paths. The same process is performed for the other two peaks 735 and 737. Consequently, the CIR duration 710 is determined as the span from the beginning to the end of the combined peak windows 733, 735, and 737. That is, the duration 710 is the span between the minimum time of the span 733 and the maximum time of the span 737.

[0059] Another aspect of the preferred receiver can be used to improve the initial channel estimate and generate an intermediate channel estimation waveform 760. Although this improvement in the channel estimate has insufficient accuracy for high-performance equalization, it can be useful for convergence of time-domain channel estimation strategies.

[0060] The intermediate channel estimation waveform 760 requires the CIR selection module 662 to select paths that are Fig. 7, marked with (X), in the initial CIR estimate (curve 740) and other points in between. A preferred strategy for such identification is found in matching pursuit methods that respond to the initial CIR estimate to partially improve the initial CIR estimate. Applying matching pursuit additionally enables improved characterization of the CIR duration 710, thereby improving the efficiency of the CIR estimation module 670.

[0061] To calculate an intermediate CIR, the following procedure can be implemented: where italics in Proc.iCIR denote scalar variables and otherwise vectors have predefined length for estimation. The P-matrix is commonly referred to in the art as the “dictionary” for reconstruction and in the present application it consists of the first L rows of the FFT matrix, and the columns are chosen as those of the locations for the pilots. The notation P(:,g) specifies the g -teColumn of the P-matrix. Although the calculations from step 3 to step 6 in Proc.iCIR are in principle the matching pursuit strategy, the stopping criterion is application-dependent. In the preferred embodiment, to obtain the intermediate CIR, matching pursuit identifies the paths and thus this process can be stopped, for example, using the criterion specified in step 7 in Proc.iCIR. Effectively, the stopping criterion measures the current peak-to-mean ratio for the most recently identified path compared to a threshold. In the example waveforms in Fig. 7 could thus stop Proc.iCIR after a minimum of 3 iterations, and the threshold for stopping the procedure is estimated for network SNR conditions as tested in a simulation environment. In certain preferred implementations, the SNR can be estimated from the initial CIR estimate. Those skilled in the art can evaluate advantageous strategies for stopping threshold determination in a given standard and implementation.

[0062] The estimation module 670 preferably responds to the intermediate channel estimation waveform 760. Such improvements from the initial channel estimate 740 that are truncated to fit within the time period 710 may not be feasible, and the TDCE strategy implemented in module 670 may require a longer convergence time. Furthermore, the estimation module 670 may advantageously respond to further metrics calculated from the waveform of the CIR selection module 662, as measured by the metrics calculation module 664. These metrics may include at least the SNR and Doppler effect in the current symbol. The estimation module 670 is the same as that described with reference to Fig. 1 and may implement linear or nonlinear methods, and preferably uses highly accurate iterative methods to obtain a time-domain channel estimate. Similarly, the phase alignment 682, plugging 684, FFT 686, and frequency equalizer 690 modules are preferably the same as the corresponding modules described in Fig. 1 and discussed above.

[0063] In Fig. 6, frequency domain processing begins with the comb filter and puncturing module 620 and is performed on groupings or blocks of a predetermined number of OFDM symbols, such as those shown in Fig. 3-5 shown fourteen symbols. The frequency domain processing of the receiver of Fig. 6 is completed by the virtual pilot interpolation module 650. The output from the virtual pilot interpolation module 650 is then transformed from the frequency domain to the time domain by the inverse fast Fourier transform (IFFT). The OFDM receiver of Fig. 6 then performs time-domain processing in modules 662, 664, 670, 682, and 684 on a symbol-by-symbol basis. That is, the preferred virtual pilot interpolation module 650 uses the queued OFDM symbols 630 simultaneously to calculate virtual pilots. All other processing is applied to each symbol individually, processing one symbol at a time.

[0064] The frequency equalizer 690 uses the output of the time-domain channel estimation module 670 to determine the equalization weights for the corresponding OFDM symbol. This procedure is well known. Given the channel frequency response (CFR) of a system, the equalizer weights are calculated as the inverse of each channel frequency response at a given carrier frequency. It is therefore desirable to phase-align the time-domain channel estimate with the received symbol for effective equalization. This is accomplished by adjusting the frequency phase shift according to the time delay for the phase alignment module 682 to apply the correct phase alignment. The phase alignment module 682 is responsive to the estimated CIR for an individual OFDM symbol.Since the CIR of the time-domain channel estimator is likely to be much shorter than the OFDM symbol duration, the stuffing module 684 preferably stuffs the CIR output by the estimator 670 before the FFT 686 transforms the channel impulse response into its channel frequency response.

[0065] The stuffing module 684 preferably stuffs the channel impulse response with zeros to extend its length to equal the FFT size. The stuffing module 684 responds to the phase-aligned CIR by increasing the number of samples in the CIR by adding zeros.

[0066] The FFT module 686 is preferably responsive to the stuffed CIR from module 684 to calculate the frequency-domain channel coefficients for all active carriers in the OFDM symbol. The frequency equalizer module 690 is preferably responsive to the channel frequency response output by the FFT 686 to determine the coefficient weights on the known frequency carriers to equalize the data before the receiver decodes the data.

[0067] The present invention has been described with respect to certain preferred embodiments. Those skilled in the art will recognize that various modifications and variations could be made to the specific preferred embodiments described herein.

Claims

[1] OFDM receiver that determines a time-domain channel impulse response, the receiver comprising: a filter (620) receiving OFDM symbols (610) containing data information and actual pilot information (520, 540, 560), the filter (620) being configured to comb filter and puncture the received OFDM symbols (610) to provide punctured OFDM symbols comprising the actual pilot information (520, 540, 560) and in which data information has been punctured; a memory (630) for receiving and storing the punctured OFDM symbols; a generator (650) coupled to the memory for generating additional pilot information (510) for the OFDM symbols based on the actual pilot information (520, 540, 560) of one or more of the punctured OFDM symbols; a time-domain channel estimator (670) that processes a first punctured OFDM symbol comprising the additional pilot information (510) to generate a channel impulse response for the first punctured OFDM symbol; and a frequency equalizer (690) that equalizes the first punctured OFDM symbol based on the channel impulse response for the first punctured OFDM symbol. [2] The receiver of claim 1, wherein the generator (650) is responsive to a plurality of punctured OFDM symbols to generate additional pilot information (510) for all of the received OFDM symbols (610) that do not include actual pilot information (520, 540, 560). [3] The receiver of claim 1, wherein the first punctured OFDM symbol is part of a block of OFDM symbols and wherein the channel impulse response is an initial channel impulse response determined based on a plurality of actual and additional pilot information (510, 520, 540, 560). [4] A receiver according to claim 3, wherein the block of OFDM symbols comprises subframes of OFDM symbols and wherein the channel impulse response is generated based on the actual additional pilot information (510, 520, 540, 560) in each subframe. [5] The receiver of claim 4, wherein each subframe comprises fourteen OFDM symbols and wherein the generator (650) generates the additional pilot information (510) for each OFDM symbol in the block. [6] The receiver of claim 1, wherein the memory (630) stores a predetermined number of symbols from a block of transmitted OFDM symbols before the generator (650) determines the additional pilot information (510) for the block. [7] The receiver of claim 6, wherein the additional pilot generator (650) identifies the additional pilot information (510) based on positions of the data-carrying carriers in the block and generates the additional pilot information (520, 540, 560) based on the received pilot information (520, 540, 560). [8] Receiver according to claim 7, wherein the generator (650) generates the additional pilot information (510) by interpolation of the received pilot information (520, 540, 560). [9] The receiver of claim 1, wherein the generator (650) comprises a Wiener filter that generates the additional pilot information (510) based on values of channel samples associated with the positions of actual pilot information (520, 540, 560). [10] Receiver according to claim 9, wherein the Wiener filter uses Doppler information and signal-to-noise ratio information to generate the additional pilot information (520, 540, 560) [11] Receiver according to one of claims 1 to 10, wherein the Wiener filter implements an interpolation of three or more nearest neighboring actual pilot information (520, 540, 560) to determine an additional pilot information (510). [12] Receiver according to one of claims 1 to 11, wherein the generator (650) generates the additional pilot information (510) for positions of the data-carrying carriers using channel samples associated with the positions of actual pilot information (520, 540, 560), wherein the receiver further comprises a CIR selector (662) responsive to an initial channel impulse response for selecting a channel impulse response interval, and wherein the time-domain channel estimator (670) is responsive to the channel impulse response interval to generate a further channel impulse response. [13] The receiver of claim 12, wherein the CIR selector (662) is configured to identify channel paths to be included in the channel impulse response interval and to terminate a process of identifying the channel paths to be included based on a normalized peak value associated with a last identified channel path.

Images